The rapid growth of large language models (LLMs) with traditional centralized fine-tuning emerges as a key technique for adapting these models to domain-specific challenges, yielding privacy risks for both model and data owners. One promising solution, called offsite-tuning (OT), is proposed to address these challenges, where a weaker emulator is compressed from the original model and further fine-tuned with adapter to enhance privacy. However, the existing OT-based methods require high computational costs and lack theoretical analysis. This paper introduces a novel OT approach based on gradient-preserving compression. By analyzing the OT problem through the lens of optimization, we propose a method that selectively applies compression techniques such as rank compression and channel pruning, preserving the gradients of fine-tuned adapters while ensuring privacy. Extensive experiments demonstrate that our approach surpasses existing OT methods, both in terms of privacy protection and model performance. Our method provides a theoretical foundation for OT and offers a practical, training-free solution for offsite-tuning of large-scale LLMs.

Large Language Models (LLMs) for Generative AI have achieved remarkable progress, evolving into sophisticated and versatile tools widely adopted across various domains and applications. However, the substantial memory overhead caused by their vast number of parameters, combined with the high computational demands of the attention mechanism, poses significant challenges in achieving low latency and high throughput for LLM inference services. Recent advancements, driven by groundbreaking research, have significantly accelerated progress in this field. This paper provides a comprehensive survey of these methods, covering fundamental instance-level approaches, in-depth cluster-level strategies, and emerging scenarios. At the instance level, we review model placement, request scheduling, decoding length prediction, storage management, and the disaggregation paradigm. At the cluster level, we explore GPU cluster deployment, multi-instance load balancing, and cloud service solutions. Additionally, we discuss specific tasks, modules, and auxiliary methods in emerging scenarios. Finally, we outline potential research directions to further advance the field of LLM inference serving.

Relevance modeling between queries and items stands as a pivotal component in commercial search engines, directly affecting the user experience. Given the remarkable achievements of large language models (LLMs) in various natural language processing (NLP) tasks, LLM-based relevance modeling is gradually being adopted within industrial search systems. Nevertheless, foundational LLMs lack domain-specific knowledge and do not fully exploit the potential of in-context learning. Furthermore, structured item text remains underutilized, and there is a shortage in the supply of corresponding queries and background knowledge. We thereby propose CPRM (Continual Pre-training for Relevance Modeling), a framework designed for the continual pre-training of LLMs to address these issues. Our CPRM framework includes three modules: 1) employing both queries and multi-field item to jointly pre-train for enhancing domain knowledge, 2) applying in-context pre-training, a novel approach where LLMs are pre-trained on a sequence of related queries or items, and 3) conducting reading comprehension on items to produce associated domain knowledge and background information (e.g., generating summaries and corresponding queries) to further strengthen LLMs. Results on offline experiments and online A/B testing demonstrate that our model achieves convincing performance compared to strong baselines.

Unlike professional Business-to-Consumer (B2C) e-commerce platforms (e.g., Amazon), Consumer-to-Consumer (C2C) platforms (e.g., Facebook marketplace) are mainly targeting individual sellers who usually lack sufficient experience in e-commerce. Individual sellers often struggle to compose proper descriptions for selling products. With the recent advancement of Multimodal Large Language Models (MLLMs), we attempt to integrate such state-of-the-art generative AI technologies into the product listing process. To this end, we develop IPL, an Intelligent Product Listing tool tailored to generate descriptions using various product attributes such as category, brand, color, condition, etc. IPL enables users to compose product descriptions by merely uploading photos of the selling product. More importantly, it can imitate the content style of our C2C platform Xianyu. This is achieved by employing domain-specific instruction tuning on MLLMs, and by adopting the multi-modal Retrieval-Augmented Generation (RAG) process. A comprehensive empirical evaluation demonstrates that the underlying model of IPL significantly outperforms the base model in domain-specific tasks while producing less hallucination. IPL has been successfully deployed in our production system, where 72% of users have their published product listings based on the generated content, and those product listings are shown to have a quality score 5.6% higher than those without AI assistance.

Large Language Models (LLMs) have demonstrated an impressive capability known as In-context Learning (ICL), which enables them to acquire knowledge from textual demonstrations without the need for parameter updates.However, many studies have highlighted that the model’s performance is sensitive to the choice of demonstrations, presenting a significant challenge for practical applications where we lack prior knowledge of user queries.Consequently, we need to construct an extensive demonstration pool and incorporate external databases to assist the model, leading to considerable time and financial costs.In light of this, some recent research has shifted focus towards zero-shot ICL, aiming to reduce the model’s reliance on external information by leveraging their inherent generative capabilities. Despite the effectiveness of these approaches, the content generated by the model may be unreliable, and the generation process is time-consuming.To address these issues, we propose Demonstration Augmentation for In-context Learning (DAIL), which employs the model’s previously predicted historical samples as demonstrations for subsequent ones.DAIL brings no additional inference cost and does not rely on the model’s generative capabilities.Our experiments reveal that DAIL can significantly improve the model’s performance over direct zero-shot inference and can even outperform few-shot ICL without any external information.

In recent years, large language models (LLMs) have driven advances in natural language processing. Still, their growing scale has increased the computational burden, necessitating a balance between efficiency and performance. Low-rank compression, a promising technique, reduces non-essential parameters by decomposing weight matrices into products of two low-rank matrices. Yet, its application in LLMs has not been extensively studied. The key to low-rank compression lies in low-rank factorization and low-rank dimensions allocation. To address the challenges of low-rank compression in LLMs, we conduct empirical research on the low-rank characteristics of large models. We propose a low-rank compression method suitable for LLMs. This approach involves precise estimation of feature distributions through pooled covariance matrices and a Bayesian optimization strategy for allocating low-rank dimensions. Experiments on the LLaMA-2 models demonstrate that our method outperforms existing strong structured pruning and low-rank compression techniques in maintaining model performance at the same compression ratio.

Despite Large Language Models (LLMs) have performed impressively in various Natural Language Processing (NLP) tasks, their inherent hallucination phenomena severely challenge their credibility in complex reasoning. Combining explainable Knowledge Graphs (KGs) with LLMs is a promising path to address this issue. However, structured KGs are difficult to utilize, and how to make LLMs understand and incorporate them is a challenging topic. We thereby reorganize a more efficient structure of KGs, while designing the KG-related instruction tuning and continual pre-training strategies to enable LLMs to learn and internalize this form of representation effectively. Moreover, we construct subgraphs to further enhance the retrieval capabilities of KGs via CoT reasoning. Extensive experiments on two KGQA datasets demonstrate that our model achieves convincing performance compared to strong baselines.

Recent generative large language models (LLMs) have exhibited incredible instruction-following capabilities while keeping strong task completion ability, even without task-specific fine-tuning. Some works attribute this to the bonus of the new scaling law, in which the continuous improvement of model capacity yields emergent capabilities, e.g., reasoning and universal generalization. However, we point out that recent LLMs still show shortcut learning behavior, where the models tend to exploit spurious correlations between non-robust features and labels for prediction, which might lead to overestimating model capabilities. LLMs memorize more complex spurious correlations (i.e., task ↔ feature ↔ label) compared with that learned from previous pre-training and task-specific fine-tuning paradigm (i.e., feature ↔ label). Based on our findings, we propose FSLI, a framework for encouraging LLMs to Forget Spurious correlations and Learn from In-context information. Experiments on three tasks show that FSFI can effectively mitigate shortcut learning. Besides, we argue not to overestimate the capabilities of LLMs and conduct evaluations in more challenging and complete test scenarios.

Although pre-trained language models (PLM) have achieved great success in question answering (QA), their robustness is still insufficient to support their practical applications, especially in the face of distribution shifts. Recently, test-time adaptation (TTA) has shown great potential for solving this problem, which adapts the model to fit the test samples at test time. However, TTA sometimes causes model collapse, making almost all the model outputs incorrect, which has raised concerns about its stability and reliability. In this paper, we delve into why TTA causes model collapse and find that the imbalanced label distribution inherent in QA is the reason for it. To address this problem, we propose Anti-Collapse Fast test-time adaptation (Anti-CF), which utilizes the source model‘s output to regularize the update of the adapted model during test time. We further design an efficient side block to reduce its inference time. Extensive experiments on various distribution shift scenarios and pre-trained language models (e.g., XLM-RoBERTa, BLOOM) demonstrate that our method can achieve comparable or better results than previous TTA methods at a speed close to vanilla forward propagation, which is 1.8× to 4.4× speedup compared to previous TTA methods.

Dynamic early exit has demonstrated great potential in coping with the sharply increasing number of pre-trained language model parameters, which can achieve a good trade-off between performance and efficiency. The existing early exit paradigm relies on training parametrical internal classifiers at each intermediate layer to complete specific tasks. Based on the predictions of these internal classifiers, different methods are designed to decide when to exit. Under this circumstance, each intermediate layer takes on both generic language representation learning and task-specific feature extraction, which makes each intermediate layer struggle to balance two types of backward loss signals during training. To break this dilemma, we propose an adapter method to decouple the two distinct types of representation and further introduce a non-parametric simplex equiangular tight frame classifier (ETF) for improvement. Extensive experiments on monolingual and multilingual tasks demonstrate that our method gains significant improvements over strong PLM backbones and early exit methods.

With the development of multilingual pre-trained language models (mPLMs), zero-shot cross-lingual transfer shows great potential. To further improve the performance of cross-lingual transfer, many studies have explored representation misalignment caused by morphological differences but neglected the misalignment caused by the anisotropic distribution of contextual representations. In this work, we propose enhanced isotropy and constrained code-switching for zero-shot cross-lingual transfer to alleviate the problem of misalignment caused by the anisotropic representations and maintain syntactic structural knowledge. Extensive experiments on three zero-shot cross-lingual transfer tasks demonstrate that our method gains significant improvements over strong mPLM backbones and further improves the state-of-the-art methods.

Non-autoregressive models have been widely used for various text generation tasks to accelerate the inference process but at the cost of generation quality to some extent. To achieve a good balance between inference speedup and generation quality, iterative NAR models like CMLM and Disco are proposed. Researchers have made much follow-up progress based on them, and some recent iterative models can achieve very promising performance while maintaining significant speedup. In this paper, we give more insights into iterative NAR models by exploring the anisotropic problem, i.e., the representations of distinct predicted target tokens are similar and indiscriminative. Upon the confirmation of the anisotropic problem in iterative NAR models, we first analyze the effectiveness of the contrastive learning method and further propose the Look Neighbors strategy to enhance the learning of token representations during training. Experiments on 4 WMT datasets show that our methods consistently improve the performance as well as alleviate the anisotropic problem of the conditional masked language model, even outperforming the current SoTA result on WMT14 EN → DE.

语法纠错是自然语言处理领域的热门任务之一,其目的是将错误的句子修改为正确的句子。为了缓解中文训练语料不足的问题,本文从数据增强的角度出发,提出一种新颖的扩充和增强数据的方法。具体地,为了使模型能更好地获取不同类型和不同粒度的错误,本文首先对语法纠错中出现的错误进行了字和词粒度的分类,在此基础上提出了融合字词粒度噪声的数据增强方法,以此获得大规模且质量较高的错误数据集。基于NLPCC2018共享任务的实验结果表明,本文提出的融合字词粒度加噪方法能够显著提升模型的性能,在该数据集上达到了最优的性能。最后,本文分析了错误类型和数据规模对中文语法纠错模型性能的影响。